Using Neural Networks to Characterize the Recoverability of Overpayments on Health Insurance Claims

Kayla McMahon
University of Wisconsin-Milwaukee
MS Graduate Student

“Health insurance companies occasionally overpay healthcare providers due to complex coordination of benefits scenarios, ineligibility, payment errors, and various other reasons. Once it is confirmed that an account was overpaid, the next step is to work with the healthcare provider to recover the overpayment. In this project, we collaborate with a division of West Corporation to mathematically characterize the recoverability of overpayments from individual healthcare providers using a feedforward neural network. We also consider the use of natural language processing to extract information from notes taken by recovery specialists detailing the work that has been put into recovering the account. A framework for adding the text mining results as additional features of the overpayments is presented with the intention of future implementation.”

Committee Members: 
Profs. Istvan Lauko (Advisor); Gabriella Pinter & Bruce Wade